Re: [PATCH for-8.2 2/2] arm/kvm: convert to kvm_get_one_reg

[Gavin Shan]

On 7/24/23 18:48, Cornelia Huck wrote:
> On Mon, Jul 24 2023, Gavin Shan <gshan@redhat.com> wrote:
> > On 7/18/23 21:14, Cornelia Huck wrote:
> > > We can neaten the code by switching the callers that work on a
> > > CPUstate to the kvm_get_one_reg function.
> > >
> > > Signed-off-by: Cornelia Huck <cohuck@redhat.com>
> > > ---
> > >    target/arm/kvm.c   | 15 +++---------
> > >    target/arm/kvm64.c | 57 ++++++++++++----------------------------------
> > >    2 files changed, 18 insertions(+), 54 deletions(-)
> >
> > The replacements look good to me. However, I guess it's worty to apply
> > the same replacements for target/arm/kvm64.c since we're here?
> >
> > [gshan@gshan arm]$ pwd
> > /home/gshan/sandbox/q/target/arm
> > [gshan@gshan arm]$ git grep KVM_GET_ONE_REG
> > kvm64.c:    err = ioctl(fd, KVM_GET_ONE_REG, &idreg);
> > kvm64.c:    return ioctl(fd, KVM_GET_ONE_REG, &idreg);
> > kvm64.c:        ret = ioctl(fdarray[2], KVM_GET_ONE_REG, &reg);
>
> These are the callers that don't work on a CPUState (all in initial
> feature discovery IIRC), so they need to stay that way.

Right, All these ioctl commands are issued when CPUState isn't around. However, 
there
are two wrappers read_sys_{reg32, reg64}(). The ioctl call in 
kvm_arm_sve_get_vls()
can be replaced by read_sys_reg64(). I guess it'd better to do this in a 
separate
patch if you agree.

Thanks,
Gavin